Time: 14:15 - 15:00
Organized by UiB
Concepts are the main building blocks in description logic knowledge bases, but being logical formulas users have difficulties writing them. Therefore, automated construction or learning of description logic concepts is an important research problem. We report on recent progress towards concept learning.
In particular, we will first give an overview of three possible learning settings: supervised, unsupervised, and interactive. Then, we will exhibit a learning algorithm in Angluin’s active learning framework for learning concepts relative to ontologies for the lightweight description logic EL.